Security and Cryptography

Security and Cryptography

Security Threats

Impersonation Pretend to be someone else to gain access to information or services

Lack of secrecy Eavesdrop on data over network

Corruption Modify data over network

Break-ins Take advantage of implementation bugs

Denial of Service Flood resource to deny use from legitimate users

Three Levels of Defense

Firewalls Filtering “dangerous” traffic at a middle point in the network

Network level security (e.g. IPsec) Host-to-host encryption and authentication Can provide security without application knowledge

Application level security True end-to-end security Requires extra effort per application Libraries help, like SSL/TLS

Private Key Cryptosystems

Finite message domain M, key domain K Key k K

Known by all parties Must be secret

Encrypt: E: M × K M Plaintext mp to ciphertext mc as mc = E(mp, k)

Decrypt: D: M × K M mp = D(mc, k) = D(E(mp, k), k)

Cryptographic security Given mc, hard to determine mp or k

Given mc and mp, hard to determine k

One Time Pad

Messages n-bit strings [b1,…,bn]

Keys Random n-bit strings [k1,…,kn]

Encryption/Decryption c = E(b, k) = b k = [b1 k1, …, bn kn]

denotes exclusive or b = D(b, k) = c k = b k k = b [0, …, 0] = b

Properties Provably unbreakable if used properly Keys must be truly random must not be used too often Key same size as message

Simple Permutation Cipher

Messages n-bit strings [b1,…,bn]

Keys Permutation of n Let = -1

Encryption/Decryption E([b1,…,bn], ) = [b (1),…,b (n)]

D([b1,…,bn], ) = [b (1),…,b (n)]

Properties Cryptanalysis possible

Data Encryption Standard (DES)

History Developed by IBM, 1975 Modified slightly by NSA U.S. Government (NIST) standard, 1977

Algorithm Uses 64-bit key, really 56 bits plus 8 parity bits

16 “rounds”• 56-bit key used to generate 16 48-bit keys• Each round does substitution and permutation using 8 S-boxes

Strength Difficult to analyze Cryptanalysis believed to be exponentially difficult in number of rounds

No currently known attacks easier than brute force

But brute force is now (relatively) easy

Other Ciphers

Triple-DES DES three times

• mc = E(D(E(mp, k1), k2, k3) Effectively 112 bits Three times as slow as DES

Blowfish Developed by Bruce Schneier circa 1993 Variable key size from 32 to 448 bits Very fast on large general purpose CPUs (modern PCs)

Not very easy to implement in small hardware Advanced Encryption Standard (AES)

Selected by NIST as replacement for DES in 2001 Uses the Rijndael algorithm Keys of 128, 192 or 256 bits

Private Key Authentication

Alice wants to talk to Bob Needs to convince him of her identity

Both have private key k Naive scheme Alice Bob

Vulnerability?

“I am Alice”, x, E(x, k)

Preventing Replay Attacks

Bob can issue a challenge phrase to Alice

Alice Bob

“I am Alice”

E(x, k)

x

Key Distribution

Have network with n entities Add one more

Must generate n new keys Each other entity must securely get its new key

Big headache managing n2 keys! One solution: use a central keyserver

Needs n secret keys between entities and keyserver

Generates session keys as needed Downsides

• Only scales to single organization level• Single point of failure

Kerberos

Trivia Developed in 80’s by MIT’s Project Athena Mythic three-headed dog guarding the entrance to Hades

Uses DES, 3DES Key Distribution Center (KDC)

Central keyserver for a Kerberos domain Authentication Service (AS)

• Database of all master keys for the domain• Users’ master keys are derived from their passwords• Generates ticket-granting tickets (TGTs)

Ticket Granting Service (TGS)• Generates tickets for communication between principals

“slaves” (read only mirrors) add reliability “cross-realm” keys obtain tickets in others Kerberos domains

Kerberos Authentication Steps

Kerberos

ServerClient

TGS

TGT Service TKT

Service REQ

Kerberos Tickets

What is a ticket? Owner (Instance and Address) A key for a pair of principles A lifetime (usually ~1 day) of the key

• Clocks in a Kerberos domain must be roughly synchronized

Contains all state Encrypted for server

Ticket-granting-ticket (TGT) Obtained at beginning of session Encrypted with secret KDC key

A needs TGT

E(kA,TGS, kA), TGTA

A AS

Kerberos – A wants to talk to B

First, get ticket from TGS

Then, use the ticket

E({A,B}, kA,TGS), TGTA

E(kA,B, kA,TGS), TKTA,B

A TGS

E({A,B}, kA,B), TKTA,B

E(m, kA,B)

E(m, kA,B)

A B

Using Kerberos

kinit Get your TGT Creates file, usually stored in /tmp

klist View your current Kerberos tickets

kdestory End session, destroy all tickets

kpasswd Changes your master key stored by the AS

“Kerberized” applications kftp, ktelnet, ssh, zephyr, etc afslog uses Kerberos tickets to get AFS token

Diffie-Hellman Key Agreement

History Developed by Whitfield Diffie, Martin Hellman Published in 1976 paper “New Directions in

Cryptography” Allows negotiation of secret key over insecure

network Algorithm

Public parameters• Prime p• Generator g < p with property: n: 1np-1, k: n = gk mod p

Alice chooses random secret a, sends Bob ga

Bob chooses random secret b, sends Alice gb

Alice computes (gb)a, Bob computes (ga)b – this is the key

Difficult for eavesdropper Eve to compute gab

Diffie-Hellman Weakness

Man-in-the-Middle attack Assume Eve can intercept and modify packets

Eve intercepts ga and gb, then sends Alice and Bob gc

Now Alice uses gac, Bob uses gbc, and Eve knows both

Defense requires mutual authentication Back to key distribution problem

Public Key Cryptosystems

Keys P, S P: public, freely distributed S: secret, known only to one entity

Properties x = D(E(x,S), P) x = D(E(x,P), S) Given x, hard to determine E(x, S) Given E(x, P), hard to determine x

Using Public Key Systems

Encryption – Bob sends to Alice Bob generates and sends mc = E (mp, PA) Only Alice is able to decrypt mp = D(mc, SA)

Authentication – Alice proves her identity Bob generates and sends challenge x Alice response s = E(x, SA) Bob checks: D(s, PA) = x

Weakness – key distribution (again) If Bob gets unauthentic PA, he can be easily attacked

Cryptographic Hash Functions

Given arbitrary length m, compute constant length digest d = h(m)

Desirable properties h(m) easy to compute given m One-way: given h(m), hard to find m Weakly collision free: given h(m) and m, hard to find m’ s.t. h(m) = h(m’)

Strongly collision free: hard to find any x, y s.t. h(x) = h(y)

Example use: password database, file distribution

Common algorithms: MD5, SHA

Comparative Performances

According to Peterson and Davie MD5: 600 Mbps DES: 100 Mbps RSA: 0.1 Mbps

Digital Signatures

Alice wants to convince others that she wrote message m Computes digest d = h(m) with secure hash Signature s = SA(d)

Digital Signature Standard (DSS)

Authentication Chains

How do you trust an unknown entity? Trust hierarchies

Certificates issued by Certificate Authorities (CAs)• Certificates are signed by only one CA• Trees are usually shallow and broad• Clients only need a small number of root CAs

– Roots don’t change frequently– Can be distributed with OS, browser

• Problem– Root CAs have a lot of power– Initial distribution of root CA certificates

X.509• Certificate format standard• Global namespace: Distinguished Names (DNs)

– Not very tightly specified – usually includes an email address or domain name

Security Vulnerabilities

Security Problems in the TCP/IP Protocol Suite – Steve Bellovin - 89

Attacks on Different Layers IP Attacks ICMP Attacks Routing Attacks TCP Attacks Application Layer Attacks

Security Flaws in IP

The IP addresses are filled in by the originating host Address spoofing

Using source address for authentication r-utilities (rlogin, rsh, rhosts etc..)

InternetInternet

2.1.1.1 C

1.1.1.1 1.1.1.2A B

1.1.1.3 S

•Can A claim it is B to the server S?

•ARP Spoofing

•Can C claim it is B to the server S?

•Source Routing

Security Flaws in IP

IP fragmentation attack End hosts need to keep the fragments till all the fragments arrive

Traffic amplification attack IP allows broadcast destination

Ping Flood

Attacking System

InternetInternet

Broadcast Enabled Network

Broadcast Enabled Network

Victim System

ICMP Attacks

No authentication ICMP redirect message

Can cause the host to switch gateways Benefit of doing this?

• Man in the middle attack, sniffing ICMP destination unreachable

Can cause the host to drop connection ICMP echo request/reply Many more…

http://www.sans.org/rr/whitepapers/threats/477.php

Routing Attacks

Distance Vector Routing Announce 0 distance to all other nodes

• Blackhole traffic• Eavesdrop

Link State Routing Can claim direct link to any other routers A bit harder to attack than DV

BGP ASes can announce arbitrary prefix ASes can alter path

TCP Attacks

ClientServer

SYN xSYN y | ACK x+1

ACK y+1

TCP Layer Attacks

TCP SYN Flooding Exploit state allocated at server after initial SYN packet

Send a SYN and don’t reply with ACK Server will wait for 511 seconds for ACK

Finite queue size for incomplete connections (1024)

Once the queue is full it doesn’t accept requests

TCP Layer Attacks

TCP Session Hijack When is a TCP packet valid?

• Address/Port/Sequence Number in window How to get sequence number?

• Sniff traffic• Guess it

– Many earlier systems had predictable initial sequence number

Inject arbitrary data to the connection

TCP Layer Attacks

TCP Session Poisoning Send RST packet

• Will tear down connection Do you have to guess the exact sequence number?• Anywhere in window is fine• For 64k window it takes 64k packets to reset

• About 15 seconds for a T1 Can reset BGP connections

Application Layer Attacks

Applications don’t authenticate properly

Authentication information in clear FTP, Telnet, POP

DNS insecurity DNS poisoning DNS zone transfer

Denial of Service

Objective make a service unusable by overloading

Consume host resources TCP SYN floods ICMP ECHO (ping) floods

Consume bandwidth UDP floods ICMP floods

Crashing the victim Ping-of-Death TCP options (unused, or used incorrectly)

Forcing more computation on routers Taking long path in processing of packets

Summary

Tools for network security: Secret keys, public/private keys, digital signature

Network security needs to be addressed at different levels Better protocols, better routers, better application level features, etc.